rfc: add notification service design doc

Problem: no design currently exists for the Flux
email service as noted in flux-framework/flux-core#4435.

Add a RFC-style document detailing this.

README.md

@@ -53,6 +53,7 @@ Table of Contents
 - [41/Job Information Service](spec_41.rst)
 - [42/Subprocess Server Protocol](spec_42.rst)
 - [43/Job List Service](spec_43.rst)
+- [44/Flux Library for Adaptable Notifications](spec_44.rst)
 
 Build Instructions
 ------------------

index.rst

@@ -283,6 +283,12 @@ standard I/O management of remote processes.
 
 The Flux Job List Service provides read-only summary information for jobs.
 
+:doc:`spec_44`
+~~~~~~~~~~~~~~
+
+This specification describes the Flux service that allows users to
+receive external notifications for events in a Flux job.
+
 .. Each file must appear in a toctree
 .. toctree::
    :hidden:
@@ -328,3 +334,4 @@ The Flux Job List Service provides read-only summary information for jobs.
    spec_41
    spec_42
    spec_43
+   spec_44

spec_44.rst

@@ -0,0 +1,258 @@
+.. github display
+   GitHub is NOT the preferred viewer for this file. Please visit
+   https://flux-framework.rtfd.io/projects/flux-rfc/en/latest/spec_28.html
+
+44/Flux Library for Adaptable Notifications Version 1
+###########################################################
+
+This specification describes the Flux service that allows users to
+receive external notifications for events in a Flux job.
+
+.. list-table::
+  :widths: 25 75
+
+  * - **Name**
+    - github.com/flux-framework/rfc/spec_44.rst
+  * - **Editor**
+    - William Hobbs <[email protected]>
+  * - **State**
+    - raw
+
+Language
+********
+
+.. include:: common/language.rst
+
+Related Standards
+*****************
+
+- :doc:`spec_14`
+- :doc:`spec_21`
+- :doc:`spec_25`
+
+Background
+**********
+
+Toward the goal of supporting users who run batch jobs with variable end time
+dependent on queues, runtime, and other factors, the Flux Library for Adaptable
+Notifications (FLAN) provides event-driven functionality that sends external
+notifications of job events. 
+
+Terminology
+***********
+
+These terms may have broader meaning in other RFCs or the Flux project. To
+avoid confusion, below is a glossary of terms as they apply in this document.
+
+Notification
+  An email or other notification triggered by FLAN but whose ultimate delivery
+  is handled by an external service.
+
+Notification-enabled jobs
+  Jobs that include a jobspec attribute requesting a notification for certain
+  events in the job's life cycle. For a more detailed definition of job events,
+  refer to :doc:`spec_21`.
+
+
+Requirements
+************
+
+ - By default in a system-instance, do not notify a user of any job events.
+   Allow the user to override this default with a jobspec attribute,
+   ``system.notify``.
+ - Support notification after any event of the job, where events are defined in
+   :doc:`spec_21`.
+ - Support email for end user notification delivery.
+ - Allow for extensibility via plugins to support more end user notification 
+   delivery services, such as Slack and Mattermost. The implementation of
+   plugins for any service other than email is not a requirement.
+ - Utilize as few resources as possible in the Flux job-manager. Under no
+   circumstances will a notification block any stage or event of a Flux job.
+ - Provide configurable rate-limiting to ensure users can never be overwhelmed
+   by notifications.
+
+Implementation
+**************
+
+FLAN SHALL be implemented in two parts:
+
+The jobtap plugin
+  A shared library based on the API defined in 
+  `flux-jobtap-plugins(7) <https://flux-framework.readthedocs.io/projects/flux-core/en/latest/man7/flux-jobtap-plugins.html>`_
+  that streams the jobids of notification-enabled jobs to the Python driver.
+
+The Python driver
+  A Python process used for tracking notification-enabled jobs through the job
+  life cycle. Started by the system instance owner on the node containing the rank 
+  0 broker in a cluster, it asynchronously monitors the events of all notification-enabled
+  jobs. It attaches callbacks to certain events and sends notifications.
+
+Initial Request
+---------------
+
+After the jobtap plugin has been loaded in the job-manager, the Python driver 
+SHALL send a ``notify.enable`` streaming RPC request to the jobtap plugin 
+at initialization.
+
+The ``notify.enable`` request has no payload.
+
+At initialization the Python driver SHALL create a kvs subdirectory, ``notify``.
+Should this directory already exist, the Python driver SHALL NOT crash. The
+Python driver SHALL traverse the existing directory and record the jobids in it.
+The Python driver SHALL reconcile the jobids in the KVS with the jobids in the
+responses to the initial RPC. 
+
+Initial Responses
+-----------------
+
+The jobtap plugin SHALL keep a record of jobids for jobs that are ACTIVE and
+notification-enabled. On initialization, all of the jobids in this record
+SHALL be sent as individual responses to the Python driver.
+
+jobid
+  As defined in :doc:`spec_19`, a single jobid for a notification-enabled job.
+
+.. note::
+  The initial responses are intended to restore state should the Python driver
+  crash.
+
+Additional Responses
+--------------------
+
+The jobtap plugin SHALL continue to send responses to the initial 
+``notify.enable`` RPC request whenever notification-enabled jobs enter the
+NEW state. The jobtap plugin SHALL add these jobids to its record 
+of ACTIVE, notification-enabled jobs.
+
+For each response received by the Python driver, the driver SHALL create a 
+KVS subdirectory, ``notify.<jobid>``. In this directory the driver SHALL
+insert keys representing the job events for which users have requested a 
+notification. These keys values SHALL be empty. The key SHALL be deleted
+after the corresponding notification is sent. 
+
+The Python driver MUST then asynchronously monitor the job as it reaches
+events of interest.
+
+When the job reaches an event of interest, the Python driver SHALL 
+generate a notification and send it to the user. The Python
+driver SHALL subsequently delete the corresponding key in the KVS, 
+``notify.<jobid>.<state>``.
+
+The ``notify.<jobid>`` KVS subdirectory SHALL be deleted when the job reaches 
+an INACTIVE state. If the ``notify.<jobid>`` directory is non-empty upon 
+reaching the INACTIVE state, this indicates some notifications have been missed.
+The Python driver SHALL send a final notification to the user documenting 
+that their notification-enabled job has reached an INACTIVE state.
+
+.. note::
+  This design is intended to ensure that no double-notifications are sent upon
+  the restart of the Python script, the jobtap plugin, or the job-manager.
+
+Error Response
+--------------
+
+If an error response is returned to ``notify.enable``, this indicates that the 
+jobtap plugin is not loaded in the job-manager. The Python driver SHALL exit 
+immediately, and print an appropriate error message.
+
+Disconnect Request
+------------------
+
+If a disconnect request is received by the jobtap plugin, this indicates the
+Python driver has exited. The jobtap plugin SHALL continue to add notification-
+enabled jobs to its record as they enter the NEW state. When the Python
+driver reconnects, the jobtap plugin SHALL respond to its initial ``notify.enable``
+RPC request with a response RPC for each jobid that is notification-enabled.
+
+User Interface
+**************
+
+Users SHALL create notification-enabled jobs by specifying an attribute in their
+job's jobspec. Jobspec attributes are defined in :doc:`spec_25`
+
+Basic Use Case
+--------------
+
+Users SHALL add the following attribute to their jobspec:
+
+.. code-block:: json
+
+  {
+    "attributes": { 
+      "system": {
+        "notify": "default"
+      }
+    }
+  }
+
+The default behavior SHALL be to send a notification via email to the user
+when the job reaches the START and FINISH events.
+
+Advanced Use Cases
+------------------
+
+Only the basic use case SHALL be supported in v1.
+
+The ``system.notify`` jobspec attribute SHALL accept a dictionary containing some
+or all of the following values:
+
+.. code-block:: json
+
+  {
+    "attributes": { 
+      "system": {
+        "notify": {
+          "service": "slack",
+          "handle": "elvis",
+          "include": ["R", "eventlog", "return_code"],
+          "states": ["start", "prolog_finish"]
+        }
+      }
+    }
+  }
+
+Edge Cases
+**********
+
+These edge cases MAY be supported in FLAN v1.
+
+Restarting the job-manager
+--------------------------
+
+In the event the job-manager crashes or is shut down the Python driver SHALL exit
+immediately and log an error. 
+
+Flux does not currently support restarting with running jobs. However, on a system
+restart, all events for all ACTIVE jobs are replayed. This means that when each
+notification-enabled job reaches the NEW event, the jobtap plugin SHALL
+send a streaming RPC response and record the jobid. The Python driver, upon 
+receiving a new jobid MUST ensure that the jobid does not have
+a previous entry in the KVS. Since the KVS is reloaded on a restart, any outstanding
+notifications SHALL have corresponding keys there. If a jobid received by the Python
+driver already has a KVS subdirectory, the Python driver SHALL ignore the job's 
+event notification requests in the jobspec and only send notifications that
+correspond with the keys in the KVS. This prevents a double-notification of the user
+for the same job state on a restart of the job-manger or FLAN service.
+
+Expiration of notifications
+---------------------------
+
+In certain cases, a restart of the service may be delayed such that events of interest
+on notification-enabled jobs are long past. FLAN MAY support an "expiration" setting
+which would stop any notification from final delivery if a set amount of time had
+passed since the event.
+
+Subinstance notifications
+-------------------------
+
+Due to the recursive launch feature of Flux, users may wish to have notifications 
+for states of batch jobs that are not at the system-instance level. This MAY NOT 
+be supported in FLAN v1.
+
+Invalid jobspec attributes
+--------------------------
+
+FLAN MAY eventually provide a plugin for validating the advanced use 
+cases detailed above. In the interim, if a user tries to utilize the advanced 
+case and provide junk keys or values, FLAN SHALL defer to default mode.
+

spell.en.pws

@@ -1,13 +1,17 @@
 rst
 
 personal_ws-1.1 en 0
+webhooks
 zeroes
 sysadmin
 admin
 org
 tarball
 tarballs
 adoc
+Mattermost
+jobids
+frobnicator
 api
 env
 github